Honor, Qualcomm Partner to Advance On-Device AI in Magic8 Smartphones
- tech360.tv
- Sep 27
- 3 min read
Honor has announced its upcoming Magic8 family of smartphones will feature the new fifth-generation Snapdragon 8 Elite chipsets. This announcement occurred during the Qualcomm Snapdragon Summit 2025, highlighting Honor’s dedication to defining artificial intelligence capabilities in smartphones.
The partnership with Qualcomm aims to provide consumers with enhanced interactions, offering benefits in computing power, storage, and battery consumption. A key focus is efficiency, not only in hardware and software but also in the user experience.
The integration of large model AI features on smartphones previously required cloud connections. In 2025, there is a push to introduce as much processing as possible directly on the device, improving data usage efficiency and reducing energy requirements.
AI models are inherently large, making their size reduction crucial for smartphone memory and storage optimisation. Honor is introducing low-bit quantisation technology in the Magic8 series to address this challenge.
This compression technology significantly reduces AI model size, with Honor demonstrating a 4.4 GB model shrinking to 1.02 GB. This allows for more efficient calculations, leading to faster speeds, lower power consumption, and enhanced user privacy through on-device processing.
However, quantisation presents a potential trade-off in accuracy, which can impact the reliability of results. Himanshu Bamoria of Athina AI noted that for many common use cases, the loss of accuracy has been negligible.
Bamoria stated, "Surprisingly, for many common use cases, the loss was negligible. The quantized models maintained competitive accuracy across benchmarks like language understanding and reasoning tasks."
He added, "This suggests that, for a vast range of applications, the extra computational burden of high-precision models might not be justified."
The Magic8 series will also feature Next-Generation Hybrid Retrieval Technology for search functionalities. This system combines traditional keyword-based search with semantic search, which focuses on meaning rather than exact matches.
Meilisearch's Ilia Markov pointed out the technical complexity of integrating multiple search algorithms required for hybrid search. He noted, "Hybrid search requires integrating multiple search algorithms (e.g., semantic search using dense embeddings). This integration can be technically complex and requires deep technical understanding."
Balancing these two search approaches is crucial to deliver more relevant results to consumers, which will likely be presented in a single, merged list. Honor aims to balance user trust and accuracy in its implementation.
A practical application of these local models is Magic Colour, enabling users to re-style photos or videos with a single natural-language prompt. Honor has provided examples of prompts that utilize various new features, such as "Save the picture as a La La Land Colour template" for photo restyling and "Help me find the photos I took at the Forbidden City" for enhanced search capabilities.
Another practical application is the AI-powered Super Sampler, primarily for gaming, which balances the GPU and NPU. This “dual engine” approach upscales low-resolution games running at lower frame rates to HD resolution at up to 120 FPS.
This dual engine enhances visual quality while offering faster processing and lower power requirements compared to GPU-only upscaling techniques. Through its partnership with Qualcomm, Honor is reworking the core engineering of its smartphones, focusing on efficient, personalised, and private AI experiences for consumers.
Honor’s Magic8 smartphone series will feature new fifth-generation Snapdragon 8 Elite chipsets.
The devices will utilise low-bit quantisation technology to reduce AI model sizes for enhanced efficiency and privacy.
Next-Generation Hybrid Retrieval Technology will combine keyword-based and semantic search for improved results.
Source: FORBES
